2 min
Doctor + setup CLI onboarding
AgentGlass
Give your AI agent eyes and hands inside iOS Simulator.
AgentGlass is a local MCP server that lets coding agents boot simulators, launch apps, tap/swipe/type, inspect UI and logs, then return reliable pass/fail output. It turns "try this manually" into repeatable automation loops you can trust.
{
"tool": "tap",
"isError": true,
"structuredContent": {
"error": {
"code": "AGENTGLASS_ACCESSIBILITY_PERMISSION_REQUIRED",
"retryable": false,
"contractVersion": "2026-03-28"
}
}
}
Hands on UI
Tap, swipe, type, and press buttons
MCP-native
Works in Claude, Cursor, and Windsurf
Structured results
Stable AGENTGLASS_* error contracts
What AgentGlass is useful for
Use it when your agent can write code but cannot reliably operate the simulator on its own.
Common jobs it solves
- Reproduce a reported UI bug from natural-language steps.
- Run post-build smoke loops before opening a PR.
- Validate onboarding, login, and paywall paths with assertions.
- Export bug reports with screenshot, logs, UI tree, and outcomes.
What it actually does
- Exposes iOS Simulator tools through MCP (
list,boot,launch,tap,swipe,type,logs,ui_tree). - Returns stable machine-readable error codes and contract metadata.
- Saves reusable flows and runs pass/fail checks automatically.
- Keeps execution local on your Mac without a hosted device farm.
2-minute MCP install
Install from the release zip, run quick checks, then connect your MCP client.
./install-agentglass-mcp.command agentglass doctor agentglass setup --boot-first-available agentglass license-status agentglass-mcp
{
"mcpServers": {
"agentglass": {
"command": "/Users/you/.local/bin/agentglass-mcp",
"args": []
}
}
}
Built for MCP-first quality loops
These capabilities move AgentGlass from prototype to a sellable agent integration.
CLI doctor/setup
Fast prerequisite checks and first-run setup for simulator, permissions, and MCP readiness.
Versioned error contracts
Stable AGENTGLASS_* error codes so agents can branch reliably instead of parsing prose.
Saved reusable flows
Persist one-click flows tied to app context and run them repeatedly.
Pass/fail assertions
Define log/UI/focus checks so sessions return objective outcomes.
Bug report export
Capture screenshot, logs, UI tree, and assertion results as Markdown + JSON.
Desktop private preview
Public release is MCP-first today; the desktop cockpit is being hardened in private preview.
Community-first pricing
Simple solo pricing: one paid plan with monthly or annual billing.
Solo Monthly
$12 /month
Best for trying AgentGlass with low commitment
- Full MCP simulator tool access
- Saved flows, assertions, and report export
- No free trial; billed immediately
- Email support and product updates
Solo Annual
$99 /year
Includes a 7-day free trial, then $99/year (effective $8.25/month)
- Same product features as Solo Monthly
- 7-day free trial before first charge
- Lower effective monthly cost
- Email support and product updates
At $12/month, break-even is tiny: at a $120/hr loaded rate, AgentGlass only needs to save about 6 minutes per month to pay for itself.